def draw(self):
"""
if sample_priors = True and random_sample = True:
draw returns a random draw of a categorical distribution with parameters drawn from a Dirichlet distribution
the hyperparameters on the Dirichlet are given by the bandit's metric with laplacian smoothing
if sample_priors = False and random_sample = True:
draw returns a random draw of a categorical distribution with parameters given by the bandit's metric
if sample_priors = True and random_sample = False:
draw returns argmax(random.dirichlet((x_0 + 1, ... , x_n_arms + 1))) where x_i is the ith value returned by
the bandit's metric.
if sample_priors = False and random_sample = False:
become a purely greedy bandit with the selected arm given by argmax(metric)
:return: The numerical index of the selected arm
"""
temp = self._schedule_fn(self.total_draws)
x = array(self._metric_fn()) * temp + 1
if self.sample_priors:
pvals = random.dirichlet(x)
else:
pvals = x / sum(x)
if self.random_sample:
return argmax(random.multinomial(1, pvals=pvals))
else:
return argmax(pvals)
def bandpower(f, Pxx, fmin, fmax):
""" integrate the power spectral density between fmin and fmax
using the trapezoidal method
"""
ind_min = scipy.argmax(f > fmin) - 1
ind_max = scipy.argmax(f > fmax) - 1
return scipy.trapz(Pxx[ind_min: ind_max], f[ind_min: ind_max])
def plot_disc_policy():
#First compute policy function...==========================================
N = 500
w = sp.linspace(0,100,N)
w = w.reshape(N,1)
u = lambda c: sp.sqrt(c)
util_vec = u(w)
alpha = 0.5
alpha_util = u(alpha*w)
alpha_util_grid = sp.repeat(alpha_util,N,1)
m = 20
v = 200
f = discretelognorm(w,m,v)
VEprime = sp.zeros((N,1))
VUprime = sp.zeros((N,N))
EVUprime = sp.zeros((N,1))
psiprime = sp.ones((N,1))
gamma = 0.1
beta = 0.9
m = 15
tol = 10**-9
delta = 1+tol
it = 0
while (delta >= tol):
it += 1
psi = psiprime.copy()
arg1 = sp.repeat(sp.transpose(VEprime),N,0)
arg2 = sp.repeat(EVUprime,N,1)
arg = sp.array([arg2,arg1])
psiprime = sp.argmax(arg,axis = 0)
for j in sp.arange(0,m):
VE = VEprime.copy()
VU = VUprime.copy()
EVU = EVUprime.copy()
VEprime = util_vec + beta*((1-gamma)*VE + gamma*EVU)
arg1 = sp.repeat(sp.transpose(VE),N,0)*psiprime
arg2 = sp.repeat(EVU,N,1)*(1-psiprime)
arg = arg1+arg2
VUprime = alpha_util_grid + beta*arg
EVUprime = sp.dot(VUprime,f)
delta = sp.linalg.norm(psiprime -psi)
wr_ind = sp.argmax(sp.diff(psiprime), axis = 1)
wr = w[wr_ind]
print w[250],wr[250]
#Then plot=================================================================
plt.plot(w,psiprime[250,:])
plt.ylim([-.5,1.5])
plt.xlabel(r'$w\prime$')
plt.yticks([0,1])
plt.savefig('disc_policy.pdf')
def testPercentErrorIsSame(self): NN.pat = zip(self.trn_d['input'], self.trn_d['target']) pyb_ws = self.net.params.copy() nn = NN() nn.wi = pyb_ws[:nn.wi.size].reshape(NN.nh, NN.ni).T nn.wo = pyb_ws[nn.wi.size:].reshape(NN.no, NN.nh).T correct = 0 wrong = 0 argmax_cor = 0 argmax_wng = 0 all_aos = [] for i, x in enumerate(self.trn_d['input']): nn.activate(x) out = self.net.activate(x) # print 'ga bp trg', nn.ao, out, self.trn_d['target'][i], '++++' if not (out - self.trn_d['target'][i]).any() else '-' all_aos.append(nn.ao.copy()) if not (out - self.trn_d['target'][i]).any(): correct += 1 else: wrong += 1 if argmax(out) == argmax(self.trn_d['target'][i]): argmax_cor += 1 else: argmax_wng += 1 print 'actual', wrong, 'wrong', correct, 'correct', float(wrong) / (wrong + correct) * 100 print 'using argmax', argmax_wng, 'wrong', argmax_cor, 'correct', float(argmax_wng) / (argmax_wng + argmax_cor) * 100 argmax_perc_err = float(argmax_wng) / (argmax_wng + argmax_cor) * 100 res = nn.sumErrors() nn_perc_err = 100 - res[1] pb_nn_perc_err = percentError(self.trainer.testOnClassData(), self.trn_d['class']) self.assertAlmostEqual(nn_perc_err, pb_nn_perc_err) self.assertAlmostEqual(nn_perc_err, pb_nn_perc_err, argmax_perc_err)
def plot_REL_ERR_SU2(self,which_case):
i=0;
thermo1 = self.select[which_case][0]
thermo2 = self.select[which_case][1]
get_REL_ERR_SU2(self,which_case)
print 'Median error SU2', sp.median(self.REL_ERR)
print 'Mean error SU2', sp.mean(self.REL_ERR)
print 'Max error SU2', max(self.REL_ERR)
print 'Min error SU2', min(self.REL_ERR)
x = getattr(self.SU2[which_case],thermo1)
y = getattr(self.SU2[which_case],thermo2)
#trusted_values = sp.where(self.REL_ERR>0<0.9*max(self.REL_ERR))
self.REL_ERR = self.REL_ERR[trusted_values]
x = x[trusted_values]
y = y[trusted_values]
scat=plt.scatter(x,y,c=self.REL_ERR, s=1)
plt.grid(which='both')
scat.set_array(self.REL_ERR)
plt.colorbar(scat)
plt.xlim((min(x)*0.95,max(x)*1.05));
plt.ylim((min(y)*0.95,max(y)*1.05));
print 'x argmax %i , x_val: %f ' %(sp.argmax(self.REL_ERR),x[sp.argmax(self.REL_ERR)])
print 'y argmax %i , y_val: %f ' %(sp.argmax(self.REL_ERR),y[sp.argmax(self.REL_ERR)])
return;
def error(self, results, expected):
err = 0
for i in range(results.shape[1]):
err += self.lsexp(results[:, i]) - sp.dot(expected[:, i], results[:, i])
misclassified = sp.sum(sp.argmax(results, axis=0) != sp.argmax(expected, axis=0))
return err, misclassified
def Problem6Real():
N = 500
w = sp.linspace(0,100,N)
w = w.reshape(N,1)
u = lambda c: sp.sqrt(c)
util_vec = u(w)
alpha = 0.5
alpha_util = u(alpha*w)
alpha_util_grid = sp.repeat(alpha_util,N,1)
m = 20
v = 200
f = discretelognorm(w,m,v)
VEprime = sp.zeros((N,1))
VUprime = sp.zeros((N,N))
EVUprime = sp.zeros((N,1))
psiprime = sp.ones((N,1))
gamma = 0.1
beta = 0.9
m = 15
tol = 10**-9
delta = 1+tol
it = 0
while (delta >= tol):
it += 1
psi = psiprime.copy()
arg1 = sp.repeat(sp.transpose(VEprime),N,0)
arg2 = sp.repeat(EVUprime,N,1)
arg = sp.array([arg2,arg1])
psiprime = sp.argmax(arg,axis = 0)
for j in sp.arange(0,m):
VE = VEprime.copy()
VU = VUprime.copy()
EVU = EVUprime.copy()
VEprime = util_vec + beta*((1-gamma)*VE + gamma*EVU)
arg1 = sp.repeat(sp.transpose(VE),N,0)*psiprime
arg2 = sp.repeat(EVU,N,1)*(1-psiprime)
arg = arg1+arg2
VUprime = alpha_util_grid + beta*arg
EVUprime = sp.dot(VUprime,f)
delta = sp.linalg.norm(psiprime -psi)
#print(delta)
wr_ind = sp.argmax(sp.diff(psiprime), axis = 1)
wr = w[wr_ind]
plt.plot(w,wr)
plt.show()
return wr
def pitch(x, fs, pitchrange=[12,120], mode='corr'):
if mode=='corr':
corr = scipy.correlate(x, x, mode='full')[len(x)-1:]
corr[:int(fs/midi2hz(pitchrange[1]))] = 0
corr[int(fs/midi2hz(pitchrange[0])):] = 0
indmax = scipy.argmax(corr)
elif mode=='ceps':
y = rceps(x)
y[:int(fs/midi2hz(pitchrange[1]))] = 0
y[int(fs/midi2hz(pitchrange[0])):] = 0
indmax = scipy.argmax(y)
return hz2midi(fs/indmax)
def generate(hmm,observation_space,n_sim): A = hmm[0] B = hmm[1] pi = hmm[2] states = sp.zeros(n_sim) observations = [] states[0] = sp.argmax(sp.random.multinomial(1,pi)) observations.append(observation_space[sp.argmax(sp.random.multinomial(1,B[states[0],:]))]) for i in range(1,n_sim): states[i] = sp.argmax(sp.random.multinomial(1,A[states[i-1],:])) observations.append(observation_space[sp.argmax(sp.random.multinomial(1,B[states[i],:]))]) return states,observations
def calcError(trainer, dataset=None):
if dataset == None:
dataset = trainer.ds
dataset.reset()
out = []
targ = []
for seq in dataset._provideSequences():
trainer.module.reset()
for input, target in seq:
res = trainer.module.activate(input)
out.append(argmax(res))
targ.append(argmax(target))
return percentError(out, targ) / 100
def generateGaussianHMM(hmm,n_sim): A = hmm[0] means = hmm[1] covars = hmm[2] pi = hmm[3] states = sp.zeros(n_sim).astype(int) K = len(means[0,:]) observations = sp.zeros((n_sim,K)) states[0] = int(sp.argmax(sp.random.multinomial(1,pi))) observations[0,:] = sp.random.multivariate_normal(means[states[0],:],covars[states[0],:,:]) for i in range(1,n_sim): states[i] = int(sp.argmax(sp.random.multinomial(1,A[states[i-1],:]))) observations[i,:] = sp.random.multivariate_normal(means[states[i],:],covars[states[i],:,:]) return states,observations
def crosscorr_phase_angle(sig1, sig2, x, max_length=10000):
"""Return the cross correlation phase angle between 2 signals
Parameters
----------
sig1 : array
signal of length L
sig2 : array
another signal of length L
x : array
time axis for the signals sig1 and sig2
max_length : int, optional
Maximum length for the signals, signals are resampled otherwise.
Default is 10 000.
"""
assert len(sig1) == len(sig2) == len(x), \
"The signals don't have the same length."
sig_length = len(sig1)
# Resample if signal is too big thus slowing down correlation computation
if sig_length > max_length:
sig1, x = resample(sig1, max_length, x)
sig2 = resample(sig2, max_length)
sig_length = max_length
corr = np.correlate(sig1, sig2, mode="same")
xmean = sig_length/2
return float(argmax(corr) - xmean)/sig_length*x[-1] # *x[-1] to scale
def tag_images_with_color_value(NUM_CLUSTERS = 4, INPUT_FOLDER = './data/covers/'):
isbn = list()
cover_color = list()
files = os.listdir(INPUT_FOLDER)
for eachFile in files:
print eachFile
im = Image.open(INPUT_FOLDER + eachFile)
im = im.resize((50, 50)) # optional, to reduce time
ar = scipy.misc.fromimage(im)
shape = ar.shape
print len(shape)
if len(shape) == 2:
ar = ar.reshape(scipy.product(shape[:1]), shape[1])
else:
ar = ar.reshape(scipy.product(shape[:2]), shape[2])
# finding clusters
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
# cluster centres:\n', codes
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
index_max = scipy.argmax(counts) # find most frequent
peak = codes[index_max]
colour = ''.join(chr(c) for c in peak).encode('hex')
isbn.append(eachFile[:-4])
cover_color.append(colour)
result = zip(isbn, cover_color)
return result
def Problem3Real():
beta = 0.9
N = 1000
u = lambda c: sp.sqrt(c)
W = sp.linspace(0,1,N)
X, Y = sp.meshgrid(W,W)
Wdiff = sp.transpose(X-Y)
index = Wdiff <0
Wdiff[index] = 0
util_grid = u(Wdiff)
util_grid[index] = -10**10
Vprime = sp.zeros((N,1))
psi = sp.zeros((N,1))
delta = 1.0
tol = 10**-9
it = 0
max_iter = 500
while (delta >= tol) and (it < max_iter):
V = Vprime
it += 1;
#print(it)
val = util_grid + beta*sp.transpose(V)
Vprime = sp.amax(val, axis = 1)
Vprime = Vprime.reshape((N,1))
psi_ind = sp.argmax(val,axis = 1)
psi = W[psi_ind]
delta = sp.dot(sp.transpose(Vprime - V),Vprime-V)
return psi
def _box_cox_transform(self, verbose=False, method='standard'):
"""
Performs the Box-Cox transformation, over different ranges, picking the optimal one w. respect to normality.
"""
from scipy import stats
a = sp.array(self.values)
if method == 'standard':
vals = (a - min(a)) + 0.1 * sp.var(a)
else:
vals = a
sw_pvals = []
lambdas = sp.arange(-2.0, 2.1, 0.1)
for l in lambdas:
if l == 0:
vs = sp.log(vals)
else:
vs = ((vals ** l) - 1) / l
r = stats.shapiro(vs)
if sp.isfinite(r[0]):
pval = r[1]
else:
pval = 0.0
sw_pvals.append(pval)
i = sp.argmax(sw_pvals)
l = lambdas[i]
if l == 0:
vs = sp.log(vals)
else:
vs = ((vals ** l) - 1) / l
self._perform_transform(vs,"box_cox")
log.debug('optimal lambda was %0.1f' % l)
return True
def most_normal_transformation(self,trans_types=SUPPORTED_TRANSFORMATIONS,
perform_trans=True, verbose=False):
"""
Performs the transformation which results in most normal looking data, according to Shapiro-Wilk's test
"""
from scipy import stats
shapiro_pvals = []
for trans_type in trans_types:
if trans_type == 'most_normal':
continue
if trans_type != 'none':
if not self.transform(trans_type=trans_type):
continue
phen_vals = self.values
#print 'sp.inf in phen_vals:', sp.inf in phen_vals
if sp.inf in phen_vals:
pval = 0.0
else:
r = stats.shapiro(phen_vals)
if sp.isfinite(r[0]):
pval = r[1]
else:
pval = 0.0
shapiro_pvals.append(pval)
if trans_type != 'none':
self.revert_to_raw_values()
argmin_i = sp.argmax(shapiro_pvals)
trans_type = trans_types[argmin_i]
shapiro_pval = shapiro_pvals[argmin_i]
if perform_trans:
self.transform(trans_type=trans_type)
log.info("The most normal-looking transformation was %s, with a Shapiro-Wilk's p-value of %.2E" % \
(trans_type, shapiro_pval))
return trans_type, shapiro_pval
def getDominantColor(img_url):
if r.exists(img_url):
cache_result = r.hmget(img_url, ['r', 'g', 'b'])
return cache_result
NUM_CLUSTERS = 5
im = Image.open(StringIO.StringIO(urllib2.urlopen(img_url).read()))
img_arr = scipy.misc.fromimage(im)
img_shape = img_arr.shape
if len(img_shape) > 2:
img_arr = img_arr.reshape(scipy.product(img_shape[:2]), img_shape[2])
codes, _ = scipy.cluster.vq.kmeans(img_arr, NUM_CLUSTERS)
original_codes = codes
for low, hi in [(60, 200), (35, 230), (10, 250)]:
codes = scipy.array([code for code in codes if not (all([c < low for c in code]) or all([c > hi for c in code]))])
if not len(codes):
codes = original_codes
else:
break
vecs, _ = scipy.cluster.vq.vq(img_arr, codes)
counts, bins = scipy.histogram(vecs, len(codes))
index_max = scipy.argmax(counts)
peak = codes[index_max]
color = [int(c) for c in peak[:3]]
r.hmset(img_url, {'r':color[0], 'g':color[1], 'b':color[2]})
#r.expire(img_url, 86400)
return color
def digshc(docs, alpha, threshold, epsilon, hr_min):
# Worth nothing that this takes in plaintext documents, _not_ vectors.
shc = SimilarityHistogramClusterer(alpha, threshold, epsilon, hr_min)
for doc in docs:
shc.fit(doc)
doc_clus_map = {}
for idx, clus in enumerate(shc.formed_clusters):
for doc_id in clus.doc_ids:
doc_clus_map.setdefault(doc_id, [])
doc_clus_map[doc_id].append(idx)
labels = []
for id in sorted(doc_clus_map):
cluster_ids = doc_clus_map[id]
if len(cluster_ids) == 1:
best_cl_id = cluster_ids[0]
else:
clusters = [shc.get_cluster(cl_id) for cl_id in cluster_ids]
sims = [shc.get_cluster_sim(cl, shc.get_doc(id)) for cl in clusters]
max_i = argmax(sims)
best_cl_id = clusters[max_i].id
labels.append(best_cl_id)
return labels
def Problem1Real():
beta = 0.9;
T = 10;
N = 100;
u = lambda c: sp.sqrt(c);
W = sp.linspace(0,1,N);
X, Y = sp.meshgrid(W,W);
Wdiff = Y-X
index = Wdiff <0;
Wdiff[index] = 0;
util_grid = u(Wdiff);
util_grid[index] = -10**10;
V = sp.zeros((N,T+2));
psi = sp.zeros((N,T+1));
for k in xrange(T,-1,-1):
val = util_grid + beta*sp.tile(sp.transpose(V[:,k+1]),(N,1));
vt = sp.amax(val, axis = 1);
psi_ind = sp.argmax(val,axis = 1)
V[:,k] = vt;
psi[:,k] = W[psi_ind];
return V,psi
def get_dominant_color(image_path):
'''
Parse image and return dominant color in image.
@param image_path: Image path to parse.
@return: Return dominant color, format as hexadecimal number.
'''
# print 'reading image'
im = Image.open(image_path)
im = im.resize((150, 150)) # optional, to reduce time
ar = scipy.misc.fromimage(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2])
# print 'finding clusters'
NUM_CLUSTERS = 5
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
# print 'cluster centres:\n', codes
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
index_max = scipy.argmax(counts) # find most frequent
peak = codes[index_max]
colour = ''.join(chr(c) for c in peak).encode('hex')
# print 'most frequent is %s (#%s)' % (peak, colour)
return "#%s" % (colour[0:6])
def par_further(tc, depth):
import scipy as SP
dview = tc[:]
dview.block = True
depths = SP.array(dview.apply(further, *[depth]))
return depths[SP.argmax(depths)]
def generate_threshold_mesh(self, min_value=0.0, max_value=1.0e9):
r"""
Generates a mesh excluding all blocks below the min_value arg. Regions
that are isolated by the thresholding are also automatically removed.
"""
#
# thresholding the data and then checking for isolated clusters
self._field.threshold_data(min_value, max_value, repl=0.0)
self._field.copy_data(self)
#
adj_matrix = self._field.create_adjacency_matrix()
num_cs, cs_ids = csgraph.connected_components(csgraph=adj_matrix,
directed=False)
# only saving the largest cluster
if num_cs > 1:
cs_count = sp.zeros(num_cs, dtype=int)
for cs_num in cs_ids:
cs_count[cs_num] += 1
self.data_vector[sp.where(cs_ids != sp.argmax(cs_count))[0]] = 0.0
self.data_map = sp.reshape(self.data_vector, (self.nz, self.nx))
#
self._field.data_map = self.data_map
self._field.data_vector = sp.ravel(self.data_map)
#
# generating blocks and vertices
mask = self.data_map > 0.0
self._generate_masked_mesh(cell_mask=mask)
def classify(self, xL, xR): a1L, a1R, a2L, a2LR, a2R, a3, z1Lb, z1LRb, z1Rb, z2b, xLb, xRb = self.forward_pass(xL, xR) if self.k == 2 : classif = sp.sign(a3); else : classif = sp.argmax(a3,axis=0); return a3, classif
def classify(self, xL, xR): x = sp.vstack([xL,xR, sp.ones(xR.shape[1])]).T tmp = sp.dot(x, self.w) return tmp, sp.argmax(tmp, axis=1)
def displayData(X, theta = None):
"""Display 2D data in a nice grid"""
width = 20
rows, cols = 10, 10
out = sp.zeros((width * rows, width * cols))
rand_indices = sp.random.permutation(5000)[0:rows * cols]
counter = 0
for y in range(0, rows):
for x in range(0, cols):
start_x = x * width
start_y = y * width
out[start_x:start_x+width, start_y:start_y+width] = X[rand_indices[counter]].reshape(width, width).T
counter += 1
img = sp.misc.toimage(out)
figure = plt.figure()
axes = figure.add_subplot(111)
axes.imshow(img)
if theta is not None:
result_matrix = []
X_biased = sp.c_[sp.ones(X.shape[0]), X]
for idx in rand_indices:
result = (sp.argmax(theta.T.dot(X_biased[idx])) + 1) % 10
result_matrix.append(result)
result_matrix = sp.array(result_matrix).reshape(rows, cols).transpose()
print(result_matrix)
plt.show()
def getPredominantColor(filename):
im = Image.open(filename).convert('RGB')
# Convert to numpy array
ar = scipy.misc.fromimage(im)
# Get dimensions
shape = ar.shape
# Convert to bidimensional array of width x height rows and 3 columns (RGB)
ar = ar.reshape(scipy.product(shape[:2]), shape[2])
# Find cluster centers and their distortions
# codes contains the RGB value of the centers
codes, dist = scipy.cluster.vq.kmeans(ar.astype(float), NUM_CLUSTERS)
# Maps all the pixels in the image to their respective centers
vecs, dist = scipy.cluster.vq.vq(ar, codes)
# Counts the occurances of each color (NUM_CLUSTER different colors after the mapping)
counts, bins = scipy.histogram(vecs, len(codes))
# Find most frequent color
index_max = scipy.argmax(counts)
peak = codes[index_max]
return peak.astype(int)
def RREFscaled(mymat):
# Pdb().set_trace()
scalevect=scipy.amax(abs(mymat),1)
scaledrows=[]
for sf,row in zip(scalevect,mymat):
row=row/sf
scaledrows.append(row)
scaledmat=scipy.vstack(scaledrows)
# scaledmat=mymat
nc=scipy.shape(scaledmat)[1]
nr=scipy.shape(scaledmat)[0]
for j in range(nr-1):
# print('=====================')
# print('j='+str(j))
pivrow=scipy.argmax(abs(scaledmat[j:-1,j]))
pivrow=pivrow+j
# print('pivrow='+str(pivrow))
if pivrow!=j:
temprow=copy.copy(scaledmat[j,:])
scaledmat[j,:]=scaledmat[pivrow,:]
scaledmat[pivrow,:]=temprow
# Pdb().set_trace()
for i in range(j+1,nr):
# print('i='+str(i))
scaledmat[i,:]-=scaledmat[j,:]*(scaledmat[i,j]/scaledmat[j,j])
return scaledmat, scalevect
def testKernelCoeffs(self):
for scale in [0.35, 0.5, 0.75, 1]:
for dim in [0,1,2,3,4]:
dgK = mango.image.discrete_gaussian_kernel(sigma=scale, dim=dim, errtol=0.001)
self.assertAlmostEqual(1.0, sp.sum(dgK), 8)
mxElem = sp.argmax(dgK)
self.assertTrue(sp.all(sp.array(dgK.shape)//2 == sp.unravel_index(mxElem, dgK.shape)))
def autofocus(self,step=5000):
if self.slide.pos[2] >= 0: step = -step
self.slide.moveZ(-step/2)
z_start = self.slide.pos[2]
self.frames.fillBuffer()
self.slide.displaceZ(step)
z_frames = self.frames.getBuffer()
#sample every kth plus its lth neighbor: for k=10,l=2 sample frame 0,10,20 and 2,12,22
k = 10
l = 2
sample_ind = [ind*k for ind in range(len(z_frames)/k)]
sample_ind2 = [ind*k+l for ind in range(len(z_frames)/k)]
f = [z_frames[ind] for ind in sample_ind]
f2 = [z_frames[ind] for ind in sample_ind2]
n = len(f)
diffs = []
for i in range(n-2):
diffs.append(ImageChops.difference(f[i],f2[i]))
motion = []
for f in diffs:
f = ImageChops.multiply(f,self.curr_mask)
motion.append(ImageStat.Stat(f).sum[0])
#g = Gnuplot.Gnuplot()
#g.plot(motion)
max_frame = scipy.argmax(motion)
max_focus = (max_frame/float(n))*step + z_start
self.slide.moveZ(max_focus)
return max_focus
def getDomIMAGEColor( imName ):
# Reference:
# http://stackoverflow.com/questions/3241929/
# python-find-dominant-most-common-color-in-an-image
# number of k-means clusters
NUM_CLUSTERS = 4
# Open target image
im = imName
im = im.resize((150, 150)) # optional, to reduce time
ar = scipy.misc.fromimage(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2])
ar = ar.astype(float)
# Find clusters
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
# Find most frequent
index_max = scipy.argmax(counts)
peak = codes[index_max]
color = ''.join(chr(int(c)) for c in peak).encode('hex')
return (peak, color)
def getObjectColor(img, box, show=False):
# crop to region of interest
sub_img = img[int(box[1]):int(box[3]), int(box[0]):int(box[2]), :]
if show: plt.imshow(sub_img)
# Run kmeans with 5 clusters to get color
ar = np.asarray(sub_img)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2]).astype(float)
codes, dist = scipy.cluster.vq.kmeans(ar, 5)
vecs, dist = scipy.cluster.vq.vq(ar, codes)
counts, bins = scipy.histogram(vecs, len(codes))
index_max = scipy.argmax(counts)
peak = codes[index_max]
# Get the hex version of the color
color = binascii.hexlify(bytearray(int(c) for c in peak)).decode('ascii')
requested_colour = tuple(int(color[i:i+2], 16) for i in (0, 2, 4))
# Determine the name of the color
actual_name, closest_name = get_colour_name(requested_colour)
if actual_name is None:
return closest_name
return actual_name
def all_saved_news(folder='model'):
import glob
from string import digits
# get just the most recent news articles file (assuming date label ordering)
news = json.load(open(glob.glob(folder+'/news*.json')[-1],"r"))
# collect text data from all articles
articles, data = [], []
for source in news.keys():
for title, article in news[source].items():
# remove numbers
for d in digits: article['text'] = article['text'].replace(d,'')
data.append(article['text'])
predictions = [prediction['probability'] for prediction in article['prediction']]
articles.append({
'source':source,
'title':title,
'url':article['url'],
'prediction':article['prediction'],
'predictedLabel':article['prediction'][argmax(predictions)]['party']
})
return articles, data
def mcmc(data, iteration):
freq = list(data)
presentstate = data.pop()
for i in range(iteration):
temp = itemfreq(np.array(freq))
dic = dict(zip(temp[:, 0], temp[:, 1]))
priorarray = prior(np.array(data) - presentstate)
index = scipy.argmax(priorarray)
nxtstate = data[index]
acceptance = min(
(1, dic[nxtstate] /
dic[presentstate])) #since we assumed symmetric prior
cointoss = np.random.random_sample()
if acceptance >= cointoss:
data.append(presentstate)
presentstate = nxtstate
freq.append(nxtstate)
data.remove(nxtstate)
else:
freq.append(presentstate)
return freq
def get_main_colour(filename):
print('reading image')
im = Image.open(f'{IMG_DIR}/{filename}')
im = im.resize((100, 100)) # optional, to reduce time
ar = np.asarray(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2]).astype(float)
print('finding clusters')
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
print('cluster centres:\n', codes)
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
index_max = scipy.argmax(counts) # find most frequent
peak = codes[index_max] # colour in rgb
colour = binascii.hexlify(bytearray(int(c) for c in peak)).decode(
'ascii') # hex colour
return tuple(map(int, peak))
def fitGaussian1D(vector):
length = len(vector)
center_guess = scipy.argmax(vector)
height_guess = vector.max()
noise_guess = vector.min()
guess_params = [
noise_guess, height_guess - noise_guess, center_guess, length / 6.0
]
max_height = height_guess * 1.1
#max_width = center_region*1.1
#max_params = [0,max_height,0,0,max_width,max_width]#,360]
#use_max = [False, True, False, False, True, True]#, True]
fits = gaussfitter.onedgaussfit(scipy.arange(0, length),
vector,
params=guess_params)
return fits[0], fits[1]
def digbc(docs, threshold):
# Worth nothing that this takes in plaintext documents, _not_ vectors.
dig = DocumentIndexGraphClusterer(threshold=threshold)
for doc in docs:
dig.index_document(doc)
doc_clus_map = {}
for idx, clus in enumerate(dig.formed_clusters):
for doc_id in clus.doc_ids:
doc_clus_map.setdefault(doc_id, [])
doc_clus_map[doc_id].append(idx)
labels = []
for id in sorted(doc_clus_map):
clusters = [dig.get_cluster(cl_id) for cl_id in doc_clus_map[id]]
sims = [dig.get_cluster_sim(cl, dig.get_doc(id)) for cl in clusters]
max_i = argmax(sims)
labels.append(clusters[max_i].id)
return labels
def GetMajorColors(path):
NUM_CLUSTERS = 10
im = Image.open(path)
im = im.resize((150, 150)) # optional, to reduce time
ar = np.asarray(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2]).astype(float)
# finding clusters
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
index_max = scipy.argmax(counts) # find most frequent
peak = codes[index_max] # RGB color
colour = binascii.hexlify(bytearray(int(c) for c in peak)).decode(
'ascii') # Hexa color
return colour
def dominantColor(filename):
NUM_CLUSTERS = 3
#print ('reading image')
im = PIL.Image.fromarray(filename)
#im = PIL.Image.open(filename)
im = im.resize((25, 25)) # optional, to reduce time
ar = np.asarray(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2]).astype(float)
#print ('finding clusters')
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
#print ('cluster centres:\n', codes)
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
index_max = scipy.argmax(counts) # find most frequent
peak = codes[index_max]
#print('max_val:\n', peak)
return peak
def background_color(img, num_clusters=3):
print('reading image')
im = Image.fromarray(img).convert('LA')
# im = Image.open(path).convert('LA')
im = im.resize((150, 150)) # optional, to reduce time
ar = scipy.misc.fromimage(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2])
print('finding clusters')
codes, dist = scipy.cluster.vq.kmeans(ar.astype('double'), num_clusters)
print('cluster centres:\n', codes)
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
result = []
index_max = scipy.argmax(counts) # find most frequent
peak = codes[index_max]
result.append(peak[0])
return result
def assign_cluster(self, document):
good_clusters = []
best_similarities = []
# calculate similarities and add to similar clusters
for cluster in self.formed_clusters:
sim_to_cluster = self.get_cluster_sim(cluster, document)
# print("%.4f" % sim_to_cluster)
if sim_to_cluster > self.threshold:
good_clusters.append(cluster)
best_similarities.append(sim_to_cluster)
# if no similar cluster found, create new
if not good_clusters:
self.create_cluster(document)
else:
if self.hard:
max_i = argmax(best_similarities)
good_clusters = [good_clusters[max_i]]
for cluster in good_clusters:
self.add_doc_to_cluster(document, cluster)
def box_cox_transform(self,
values,
lambda_range=(-2.0, 2.0),
lambda_increment=0.1,
verbose=False,
method='standard'):
"""
Performs the Box-Cox transformation, over different ranges, picking the optimal one w. respect to normality.
"""
from scipy import stats
a = sp.array(values)
if method == 'standard':
vals = (a - min(a)) + 0.1 * sp.std(a)
else:
vals = a
sw_pvals = []
lambdas = sp.arange(lambda_range[0],
lambda_range[1] + lambda_increment,
lambda_increment)
for l in lambdas:
if l == 0:
vs = sp.log(vals)
else:
vs = ((vals**l) - 1) / l
r = stats.shapiro(vs)
if sp.isfinite(r[0]):
pval = r[1]
else:
pval = 0.0
sw_pvals.append(pval)
# log.info(sw_pvals)
i = sp.argmax(sw_pvals)
l = lambdas[i]
if l == 0:
vs = sp.log(vals)
else:
vs = ((vals**l) - 1) / l
# self._perform_transform(vals,"box-cox")
sys.stdout.write('optimal lambda was %0.1f\n' % l)
return vals
def getLPDataFWHM(lprof):
'''
Calculate the FWHM of the line profile.
@param lprof: the line profile object.
@type lprof: LPDataReader()
@return: the fwhm of the line
@rtype: float
'''
flux = lprof.getFlux()
vel = lprof.getVelocity()
vlsr = lprof.getVlsr()
maxval = max(flux)
i_mid = argmax(flux)
flux1 = flux[:i_mid]
flux2 = flux[i_mid:]
v1 = vel[argmin(abs(flux1 - maxval / 2.))]
v2 = vel[argmin(abs(flux2 - maxval / 2.)) + i_mid]
return v2 - v1
def test_sim1():
p = 0.6
payoutBarrier = 1.0
beta = 0.99
nPeriods = 1000
nRuns = 1000
M_grid = scipy.arange(21.0)
V_array = scipy.zeros((len(M_grid), len(M_grid)))
for (iM, M) in enumerate(M_grid):
#print("M=%f" % M)
b_grid = M_grid[M_grid <= M]
for (ib, b) in enumerate(b_grid):
V_simulated_list = []
for j in range(nRuns):
firm = Firm(M, beta, b)
(survived, age, V) = simulateFirm(firm, p, nPeriods)
V_simulated_list.append(V)
EV = scipy.mean(V_simulated_list)
V_array[iM, ib] = EV
for (iM, M) in enumerate(M_grid):
V_of_b = V_array[iM,:]
print(V_of_b)
optimal_ib = scipy.argmax(V_of_b)
optimal_b = M_grid[optimal_ib]
print("for M=%f, optimal b is %f" % (M, optimal_b))
fig = plt.figure()
ax = Axes3D(fig)
vals = V_array.flatten()
[xlist, ylist] = zip(*itertools.product(M_grid, b_grid))
ax.scatter(xlist, ylist, vals)
ax.set_xlabel('M')
ax.set_ylabel('b')
ax.set_zlabel('EV')
def get_dominant_hexcolor(image: Image) -> str:
resized_image = image.resize((200, 200))
pixel_array = np.asarray(resized_image)
pixel_array_shape = pixel_array.shape
reshaped_pixel_array = pixel_array.reshape(
scipy.product(pixel_array_shape[:2]),
pixel_array_shape[2]).astype(float)
centroids, _ = scipy.cluster.vq.kmeans(reshaped_pixel_array,
3,
iter=20,
thresh=1e-5)
codes, _ = scipy.cluster.vq.vq(reshaped_pixel_array, centroids)
color_counts, _ = scipy.histogram(codes, len(centroids))
highest_index = scipy.argmax(color_counts)
peak = centroids[highest_index]
hex_color = "#" + binascii.hexlify(bytearray(
int(c) for c in peak)).decode('ascii')
return hex_color
def box_cox_transform(self, pid, lambda_range=(-2.0, 2.0), lambda_increment=0.1, verbose=False, method='standard'):
"""
Performs the Box-Cox transformation, over different ranges, picking the optimal one w. respect to normality.
"""
from scipy import stats
a = sp.array(self.phen_dict[pid]['values'])
if method == 'standard':
vals = (a - min(a)) + 0.1 * sp.std(a)
else:
vals = a
sw_pvals = []
lambdas = sp.arange(lambda_range[0], lambda_range[1] + lambda_increment, lambda_increment)
for l in lambdas:
if l == 0:
vs = sp.log(vals)
else:
vs = ((vals ** l) - 1) / l
r = stats.shapiro(vs)
if sp.isfinite(r[0]):
pval = r[1]
else:
pval = 0.0
sw_pvals.append(pval)
print sw_pvals
i = sp.argmax(sw_pvals)
l = lambdas[i]
if l == 0:
vs = sp.log(vals)
else:
vs = ((vals ** l) - 1) / l
if not self.phen_dict[pid]['transformation']:
self.phen_dict[pid]['raw_values'] = self.phen_dict[pid]['values']
self.phen_dict[pid]['transformation'] = 'box-cox'
else:
self.phen_dict[pid]['transformation'] = 'box-cox(' + self.phen_dict[pid]['transformation'] + ')'
self.phen_dict[pid]['values'] = vs.tolist()
if verbose:
print 'optimal lambda was %0.1f' % l
return True
def most_normal_transformation(
self,
pid,
trans_types=['none', 'sqrt', 'log', 'sqr', 'exp', 'arcsin_sqrt'],
perform_trans=True,
verbose=False):
"""
Performs the transformation which results in most normal looking data, according to Shapiro-Wilk's test
"""
#raw_values = self.phen_dict[pid]['values']
from scipy import stats
shapiro_pvals = []
for trans_type in trans_types:
if trans_type != 'none':
if not self.transform(pid, trans_type=trans_type):
continue
phen_vals = self.get_values(pid)
#print 'sp.inf in phen_vals:', sp.inf in phen_vals
if sp.inf in phen_vals:
pval = 0.0
else:
r = stats.shapiro(phen_vals)
if sp.isfinite(r[0]):
pval = r[1]
else:
pval = 0.0
shapiro_pvals.append(pval)
#self.phen_dict[pid]['values'] = raw_values
if trans_type != 'none':
self.revert_to_raw_values(pid)
argmin_i = sp.argmax(shapiro_pvals)
trans_type = trans_types[argmin_i]
shapiro_pval = shapiro_pvals[argmin_i]
if perform_trans:
self.transform(pid, trans_type=trans_type)
if verbose:
print "The most normal-looking transformation was %s, with a Shapiro-Wilk's p-value of %0.6f" % \
(trans_type, shapiro_pval)
return trans_type, shapiro_pval
def main(npImage):
#print('reading image')
#im = Image.open('image.jpg')
im = Image.fromarray(npImage)
im = im.resize((150, 150)) # optional, to reduce time
ar = np.asarray(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2]).astype(float)
#print('finding clusters')
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
#print('cluster centres:\n', codes)
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
index_max = scipy.argmax(counts) # find most frequent
peak = codes[index_max]
# colour = binascii.hexlify(bytearray(int(c) for c in peak)).decode('ascii')
# print('most frequent is %s (#%s)' % (peak, colour))
peak = [peak[2], peak[1], peak[0]]
return peak
def oneGeneration(self):
if self.generation > 0:
self.calculateAverageFitness()
# evaluate fitness
self.fitnesses = []
for indiv in self.currentpop:
self.fitnesses.append(self.targetfun(indiv))
# determine the best values
best = argmax(array(self.fitnesses))
self.bestfitness = self.fitnesses[best]
self.bestx = self.currentpop[best]
self.allgenerations.append((self.currentpop, self.fitnesses))
if self.fitnessSmoothing:
self._smoothFitnesses()
# selection
tmp = zip(self.fitnesses, self.currentpop)
tmp.sort(key = lambda x: x[0])
tmp2 = list(reversed(tmp))[:self.selectionSize()]
parents, self.parentFitnesses = map(lambda x: x[1], tmp2), map(lambda x: x[0], tmp2)
self.currentpop = self.crossOver(parents, self.popsize)
# add one random offspring
#self.currentpop[-1] = randn(self.xdim)
#self.crossovervectors[-1] = [0]*self.xdim
for child in self.currentpop:
self.mutate(child)
if self.generation > 0:
self.updateLinkageMatrix()
if self.verbose:# and self.generation % 10 == 0:
# TODO: more extensive output
print self.rawlm
def run_model1_qa(objects_filename, flickr_cache):
#load a location file
loc_info = load_location_file(objects_filename)
#create a likelihood map
l_map = model1_qanswer(flickr_cache, loc_info)
olists, igains, p_ts, p_fs = [], [], [], []
for k in range(5):
olist, igain, p_t, p_f = l_map.maximal_information_gain()
olists.append(olist)
igains.append(igain)
p_ts.append(p_t)
p_fs.append(p_f)
for i in range(len(igain)):
print "obj:", olist[i], " I:", igain[i], " P(t):", p_t[
i], " P(f):", p_f[i]
if (len(igain) == 0):
j = -1
else:
j = argmax(igain)
if (not j == -1):
print "adding allowed object:", olist[j]
l_map.add_allowed_object(olist[j])
print "best=", olist[j]
res = {}
res['object_lists'] = olists
res['information_gain'] = igains
res['prob_true'] = p_ts
res['prob_false'] = p_fs
cPickle.dump(res, open('m1_qanswer.pck', 'w'))
def plot_eigvect(vect, labels=None, bottom=0, num_label=5, label_offset=0.15):
"""
Plot a given eigenvector.
If a list of labels is passed in, the largest (in magnitude) num_label bars
will be labeled on the plot. label_offset controls how much the labels
are shifted from the top of the bars for clarity.
bottom controls where the bar plot is centered along the y axis. This is
useful for plotting several e'vectors on the same axes.
"""
# The 0.4 centers the bars on their numbers, accounting for the default
# bar width of 0.8
vect = scipy.real(vect)
max_index = scipy.argmax(abs(vect))
if vect[max_index] < 0:
vect = -vect
bar(scipy.arange(len(vect)) - 0.4,
vect / scipy.linalg.norm(vect),
bottom=bottom)
a = list(axis())
a[0:2] = [-.03 * len(vect) - 0.4, (len(vect) - 1) * 1.03 + 0.4]
if labels is not None:
mags = zip(abs(vect), range(len(vect)), vect)
mags.sort()
mags.reverse()
for mag, index, val in mags[:num_label]:
name = labels[index]
text(index,
val + scipy.sign(val) * label_offset,
name,
horizontalalignment='center',
verticalalignment='center')
a[2] -= 0.1
a[3] += 0.1
axis(a)
def Problem3Real():
beta = 0.95
N = 1000
u = lambda c: sp.sqrt(c)
W = sp.linspace(0,1,N)
W = W.reshape(N,1)
X, Y = sp.meshgrid(W,W)
Wdiff = Y-X
index = Wdiff <0
Wdiff[index] = 0
util_grid = u(Wdiff)
V = sp.zeros((N,1))
z = 0
r = 15
delta =1
while (delta > 10**-9):
z += 1
#print(z)
#Update Policy Function
arg = util_grid + beta*sp.transpose(V)
arg[index] = -10**10
psi_ind = sp.argmax(arg,axis = 1)
V_prev = V
#Iterate on Value Function
for j in sp.arange(0,r):
V = u(W-W[psi_ind]) + beta*V[psi_ind]
delta = sp.dot(sp.transpose(V_prev - V),V_prev-V)
return W[psi_ind]
def Problem2Real():
beta = 0.95
N = 1000
u = lambda c: sp.sqrt(c)
psi_ind = sp.arange(0,N)
W = sp.linspace(0,1,N)
X, Y = sp.meshgrid(W,W)
Wdiff = Y-X
index = Wdiff <0
Wdiff[index] = 0
util_grid = u(Wdiff)
I = sp.sparse.identity(N)
delta = 1
z = 0
while (delta > 10**-9):
z = z+1
#print(z)
psi_prev = psi_ind.copy()
rows = sp.arange(0,N)
columns = psi_ind
data = sp.ones(N)
Q = sp.sparse.coo_matrix((data,(rows,columns)),shape = (N,N))
Q = Q.tocsr()
#Solve for Value Function
V = spsolve(I-beta*Q,u(W-W[psi_ind]))
#Find Policy Function
arg = util_grid + beta*V
arg[index] = -10**10
psi_ind = sp.argmax(arg,axis = 1)
delta = sp.amax(sp.absolute(W[psi_ind]-W[psi_prev]))
return W[psi_ind]
def Problem1Real():
beta = 0.9
T = 10
N = 100
u = lambda c: sp.sqrt(c)
W = sp.linspace(0, 1, N)
X, Y = sp.meshgrid(W, W)
Wdiff = Y - X
index = Wdiff < 0
Wdiff[index] = 0
util_grid = u(Wdiff)
util_grid[index] = -10**10
V = sp.zeros((N, T + 2))
psi = sp.zeros((N, T + 1))
for k in xrange(T, -1, -1):
val = util_grid + beta * sp.tile(sp.transpose(V[:, k + 1]), (N, 1))
vt = sp.amax(val, axis=1)
psi_ind = sp.argmax(val, axis=1)
V[:, k] = vt
psi[:, k] = W[psi_ind]
return V, psi
def createPattern(image):
colorArray = []
ar = np.asarray(image)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2]).astype(float)
codes, dist = scipy.cluster.vq.kmeans(ar, 5) #Baskın 5 rengin kümelenmesi.
print("codes: ", codes, " uzaklık: ", dist)
vecs, dist = scipy.cluster.vq.vq(ar, codes)
print("vektör: ", vecs, " uzaklik: ", dist)
counts, bins = scipy.histogram(vecs, len(codes))
print("counts: ", counts, " bins: ", bins)
index_max = scipy.argmax(
counts) #En sık kullanılan 1 rengin indexini bulur
print("indexmaks: ", index_max)
peak = codes[index_max]
print(peak)
for i in range(0, 5):
colorArray.append(codes[i].astype("uint8").tolist())
colorArray.append(peak.astype("uint8").tolist()) #En baskın renk
print("colorArray")
return colorArray
def getDomColor(imgFileName):
# Reference:
# http://stackoverflow.com/questions/3241929/
# python-find-dominant-most-common-color-in-an-image
# number of k-means clusters
NUM_CLUSTERS = 4
# Open target image
im = Image.open(imgFileName)
im = im.resize((150, 150)) # optional, to reduce time
ar = scipy.misc.fromimage(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2])
ar = ar.astype(float)
# Find clusters
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
# print 'cluster centres:\n', codes
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
# Find most frequent
index_max = scipy.argmax(counts)
peak = codes[index_max]
color = ''.join(chr(int(c)) for c in peak).encode('hex')
#print 'most frequent is %s (#%s)' % (peak, color)
# Extra Bonus coolbeans: save image using only the N most common colors
# c = ar.copy()
# for i, code in enumerate(codes):
# c[scipy.r_[scipy.where(vecs==i)],:] = code
# scipy.misc.imsave(imgFileName[:-4] + "CLUS" + ".jpg", c.reshape(*shape))
# print 'saved clustered image'
return (peak, color)
def PLU(A):
# throw warning flag when the number is too small
# (close to 0)
ok = 1
small = 1e-12
n = scipy.shape(A)[0]
U = copy.copy(A)
L = scipy.identity(n)
P = scipy.identity(n)
for j in range(1, n):
print(j, " operation")
print("This is U")
print(U)
print("This is L")
print(L)
print("This is P")
print(P)
print()
s = scipy.argmax(abs(U[j - 1:n, j - 1])) + j - 1
# argmax returs the index of that number
if s != j - 1:
U = swap(U, s, j - 1, n)
P = swap(P, s, j - 1, n)
if j > 1:
L = swap(L, s, j - 1, j - 1)
for i in range(j + 1, n + 1):
if abs(U[j - 1, j - 1]) < small:
print("Near-zero pivot!")
ok = 0
break
L[i - 1, j - 1] = U[i - 1, j - 1] / U[j - 1, j - 1]
for k in range(j, n + 1):
U[i - 1,
k - 1] = U[i - 1, k - 1] - L[i - 1, j - 1] * U[j - 1, k - 1]
return L, U, P, ok
def parse_plink_snps(genotype_file, snp_indices):
plinkf = plinkfile.PlinkFile(genotype_file)
samples = plinkf.get_samples()
num_individs = len(samples)
num_snps = len(snp_indices)
raw_snps = sp.empty((num_snps, num_individs), dtype='int8')
# If these indices are not in order then we place them in the right place while parsing SNPs.
snp_order = sp.argsort(snp_indices)
ordered_snp_indices = list(snp_indices[snp_order])
ordered_snp_indices.reverse()
# Iterating over file to load SNPs
snp_i = 0
next_i = ordered_snp_indices.pop()
line_i = 0
max_i = ordered_snp_indices[0]
while line_i <= max_i:
if line_i < next_i:
next(plinkf)
elif line_i == next_i:
line = next(plinkf)
snp = sp.array(line, dtype='int8')
bin_counts = line.allele_counts()
if bin_counts[-1] > 0:
mode_v = sp.argmax(bin_counts[:2])
snp[snp == 3] = mode_v
s_i = snp_order[snp_i]
raw_snps[s_i] = snp
if line_i < max_i:
next_i = ordered_snp_indices.pop()
snp_i += 1
line_i += 1
plinkf.close()
assert snp_i == len(raw_snps), 'Parsing SNPs from plink file failed.'
num_indivs = len(raw_snps[0])
freqs = sp.sum(raw_snps, 1, dtype='float32') / (2 * float(num_indivs))
return raw_snps, freqs
def PLU(A):
ok = 1
small = 1e-12
n = scipy.shape(A)[0]
U = copy.copy(A)
L = scipy.identity(n)
P = scipy.identity(n)
for j in range(1, n):
s = scipy.argmax(abs(U[j - 1:n, j - 1])) + j - 1
if s != j - 1:
U = swap(U, s, j - 1, n)
P = swap(P, s, j - 1, n)
if j > 1:
L = swap(L, s, j - 1, j - 1)
for i in range(j + 1, n + 1):
if abs(U[j - 1, j - 1]) < small:
print("Near-zero pivot!")
ok = 0
break
L[i - 1, j - 1] = U[i - 1, j - 1] / U[j - 1, j - 1]
for k in range(j, n + 1):
U[i - 1,
k - 1] = U[i - 1, k - 1] - L[i - 1, j - 1] * U[j - 1, j - 1]
return L, U, P, ok
def get_main_color(url):
NUM_CLUSTERS = 5
# print('reading image')
response = requests.get(url)
im = Image.open(BytesIO(response.content))
im = im.resize((150, 150)) # optional, to reduce time
ar = np.asarray(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2]).astype(float)
# print('finding clusters')
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
# print('cluster centres:\n', codes)
vecs, dist = scipy.cluster.vq.vq(ar, codes) # assign codes
counts, bins = scipy.histogram(vecs, len(codes)) # count occurrences
index_max = scipy.argmax(counts) # find most frequent
peak = codes[index_max]
# print(peak)
colour = binascii.hexlify(bytearray(int(c) for c in peak)).decode('ascii')
# print(colour)
return peak, colour
def find_dominant_color(imageurl: str, local=False):
try:
NUM_CLUSTERS = 10
if (local):
im = Image.open(imageurl)
else:
im = Image.open(requests.get(imageurl, stream=True).raw)
im = im.resize((25, 25))
ar = np.asarray(im)
shape = ar.shape
ar = ar.reshape(scipy.product(shape[:2]), shape[2]).astype(float)
codes, dist = scipy.cluster.vq.kmeans(ar, NUM_CLUSTERS)
vecs, dist = scipy.cluster.vq.vq(ar, codes)
counts, bins = scipy.histogram(vecs, len(codes))
index_max = scipy.argmax(counts)
peak = codes[index_max]
colour = binascii.hexlify(bytearray(int(c)
for c in peak)).decode('ascii')
try:
return int(hex(int(colour, 16))[:8], 0)
except:
return 0xffffff
except IndexError:
return 0xffffff
